Human speech is typically a spontaneous interaction between talkers, but existing techniques for observing the underlying speech movements crucial for shaping the sounds of language are mostly limited to highly controlled studies of single speakers. In this proposal we pioneer a new approach that for the first time supports simultaneous direct observation of speech articulation by two face-to-face talkers, and apply it to studying the mutual adaptation that occurs between them as they interact in conversation. It is known from previous work that when speakers converse with one another, they typically make adjustments in their patterns of speech. When such adjustments are in the direction of increased similarity, the effect is known as speaker alignment or phonetic convergence. Until now this phenomenon has been studied primarily through the analysis of conversational audio recordings (including subjective ratings by listeners), but the skilled movements of the speech articulators that underlie such audio have not been accessible. This new approach, exploiting two synchronized electromagnetic articulometer (EMA) systems, will for the first time provide direct observation of the kinematic aspects of phonetic convergence. Because such convergence is a fundamental means of facilitating communication and establishing social distance between speakers, baseline knowledge of its operation is important for suggesting remedies when it fails to operate as expected, as for example with persistent foreign accent. As a practical application this proposal will compare the extent of kinematic convergence observed between pairs of native English speakers, to that found between mixed pairs with one native and one non-native speaker. This will lead to improved understanding of the articulatory aspects of foreign accent, and suggest possibilities for improving methods of speech remediation and second language instruction.